Process Mining for Optimization of a Loan Approval Process in a Financial Institution

摘要

Financial institutions are experiencing drastic changes after the global financial crisis of 2008. On the one hand, financial institutions need to quickly adapt to new compliance regulations that require adapting their internal processes. On the other hand, Fintechs are changing the traditional banking rules by introducing innovative financial processes.udTherefore, financial institutions need to be able to improve their operational inefficiencies, and represents a business case for the application of process mining techniques. In this final degree project a real case is faced through different process and data mining techniques.udThe data used was provided in the BPIC 2017. This challenge proposes a real use case where a Dutch financial institution provides event logs of the loan approval process, with 1.202.267 events pertaining to 31.509 applications. For the approach, we leverage the Disco process mining tool with Python-based data analysis and visualization tools such as Pandas in order to combine different granularity inspection techniques to provide answers to the given questions.udIn particular, we focus on the main requests from the BPIC 2017 challenge, which are: throughput times per part of the process, influence on the frequency of incompleteness to the final outcome and the frequency of customers asking for more than one offer. Our approach has consisted in identifying the process phases and analyzing each question by phase and then globally. Finally, we discuss concluding remarks and future work.